Towards Automated Fracture Detection: A Study on Feature-Based Transfer Learning with Xception for Hairline and Avulsion Fractures

Description

Accurate diagnosis of bone fractures, particularly avulsion and hairline fractures, is essential for effective treatment and recovery. Traditional methods relying on radiologists’ expertise can be subjective and prone to errors. This study investi-gates the application of feature-based transfer learning with the Xception model to enhance fracture classification in X-ray images. A dataset of 80 hairline fractures, 100 avulsion fractures and 50 non-fracture images were used. Pre-trained Xcep-tion convolutional neural network (CNN) models extracted discriminative fea-tures, which were then classified using Support Vector Machine (SVM), Logistic Regression (LR), and k-Nearest Neighbors (kNN). The results demonstrated that both SVM and LR achieved high accuracy, with SVM showing superior general-ization due to its ability to handle complex, non-linear patterns. LR exhibited reli-able performance but faced challenges with non-linear boundaries, while kNN was sensitive to noise and parameter selection. Despite these challenges, the study confirms that feature-based transfer learning improves classification effi-ciency and accuracy compared to training CNNs from scratch. These findings highlight the potential of integrating deep learning and machine learning for de-veloping automated fracture detection systems to assist healthcare professionals. Future work should explore advanced architectures and refine model parameters to further enhance performance. This study lays a foundation for improving diag-nostic accuracy in medical imaging, contributing to better patient care and out-comes.

Period	22 Aug 2024
Event title	International Conference on Intelligent Manufacturing and Robotics 2024
Event type	Conference
Location	Taicang, Suzhou, ChinaShow on map
Degree of Recognition	International